Tell Them How

There's a line I keep seeing on service business websites and it makes me wince every time.

"All of our content is written by real humans."

I get why it's there. I understand the instinct. But read it as a customer for a second. What that sentence actually communicates is that the company thinks its audience is suspicious, that the suspicion is worth a defensive statement, and that the most interesting thing about their work is what they didn't use to make it.

Nobody in the history of commerce has bought anything because of a tool that wasn't involved. "Hand written by humans" is not a value proposition. It's an alibi.

And it's about to get worse, because that sentence is one detector result away from being a liability. If any part of the process touched a model, and increasingly some part of it did, that line goes from defensive to false. Not morally, technically. But try explaining the difference to somebody who just got a percentage back from a tool they don't understand.

Meanwhile the marking infrastructure keeps getting more real. Text watermarks since the start of this month. Signed provenance metadata on files. C2PA in cameras and in Adobe and now in AI outputs. The direction of travel is obvious: over the next couple of years, the origin of a piece of content becomes visible by default rather than debatable.

So the question stops being whether people find out. They find out. The question is whether they find out from you, in a sentence you wrote, or from a browser extension in the middle of deciding whether to hire you.

Almost everyone lands on one of two positions, and both are worse than the thing in the middle.

Position one is silence. Say nothing, hope nobody asks. This is the default and it's not crazy, but it has a specific failure mode. Silence is fine until the moment it isn't, and then it reads as concealment, because from the outside a deliberate non answer and a guilty non answer look identical.

The cost isn't a scandal. It's smaller and more constant than that. It's the prospect who wonders, doesn't ask, and quietly picks someone else. You never see that. It just shows up as a conversion rate you assume is normal.

Position two is the badge. "Powered by AI." "AI enhanced workflow." "We leverage cutting edge artificial intelligence."

This one's worse. It's 2026. Announcing you use AI is like announcing you use email. It signals nothing about quality and quite a lot about how recently you got excited about something. Worse, it invites the exact question you don't want, which is: so what am I paying you for.

The middle position is boring and it works. You describe your process, briefly, like a person who has thought about it. One sentence, maybe two. Not a policy page. Not a manifesto. A line.

Here's the mechanic, and once you see it you can't unsee it.

A disclosure done badly answers "did you use AI." A disclosure done well answers "what do I get from you that I can't get from the machine directly."

Same sentence. Completely different job.

Because that second question is the one your prospect actually has, and it's the one they can't ask politely. They know they could open a chat window and get a draft. What they don't know is what you add on top of that, and whether it's worth what you charge.

When you describe your process, you're not confessing. You're itemizing your judgment.

Watch the difference.

Weak: "We use AI tools to help create our content."

Better: "First drafts come out of a model. Then I rewrite them against six years of what actually converted for clients in this industry, which is the part you're paying for."

The second one mentions AI more specifically and comes out ahead, because the sentence is about the six years, not the model. The tool is the setup. Your judgment is the punchline.

That's the whole trick. Name the tool, then name what you do after it, and make the second part longer.

Four places, in order of how much they matter.

Your services or about page. One line, in your voice, near where you describe how you work. Not a footer, not a legal page, not a modal nobody opens. If you publish on beehiiv or anything similar, that means the about page and the sign up page, not a policy link in the footer. It belongs in the flow of the sentence where you're already explaining your process, because that's the honest place for it and honest placement is most of the signal.

Your proposals. This is the highest value spot and almost nobody does it. A single line in the methodology section, before anybody asks, does three things at once. It closes a question the buyer had. It makes you look like someone who's thought about the thing they're nervous about. And it inoculates you against the awkward email in month four.

I've watched proposals win on this. Not because the disclosure was impressive, but because it was the only one in the stack that named the elephant, and the buyer read that as competence.

Your published content, when it's relevant. Not every blog post needs a disclaimer, please don't do that. But when the topic is adjacent, or when you're showing work, or when the piece itself is an example of the thing you sell, a line in the author bio or a sentence in the piece is worth having.

Your client agreements. Different register, same idea. This one's less about trust and more about not having a bad conversation later. Short clause, plain language, states what assistance you use and what you keep as record. Your lawyer will write it in a paragraph and it'll cost you less than a nice dinner.

Twenty minutes. Do it properly once and it lasts for years.

Start by writing down what you actually do. Honestly, for yourself, not for publication. Where does the model come in. What do you do before it. What do you do after. Be specific enough that it would be embarrassing if it weren't true.

Most people are surprised here. They discover the model is doing less than they thought, and it's doing it in a narrower window than they thought, usually somewhere in the middle of the process. The research is theirs. The strategy is theirs. The final judgment is theirs. The model wrote a scaffold in the middle that got mostly torn out.

That's a good story. It's also, for most competent operators, the true one. You just never said it out loud because you assumed it sounded like an admission.

Then compress it. One sentence naming the tool's role. One sentence naming yours. Keep the second one concrete, because concrete is what does the work.

"Drafts start with an assistant" is fine. "Then everything gets rewritten against the eleven questions our clients actually ask on sales calls" is the part that sells.

Then read it out loud. If it sounds defensive, cut the defensiveness. If it sounds like a press release, cut the press release. If it sounds like something you'd say to a client across a table, you're done.

One more thing, and this is the one people get wrong. Don't overclaim the human part. If a model wrote most of a piece and you edited lightly, don't write a sentence implying you agonized over every word. That's the version that blows up, because it's the version that's false, and falseness is the only thing here that actually costs you.

If your honest answer is "the model does a lot and my value is in selection and judgment rather than sentences," then write that. It's a real position and there are clients who want exactly it. What you cannot do is claim craft you didn't apply, because that claim is the one somebody eventually tests.

Abstract advice is easy to agree with and hard to use, so here are four written out. Steal the structure, replace the specifics.

Consultant or agency. "Research and first drafts run through AI. Everything after that is me, working from the client interviews and eleven years of watching which recommendations actually got implemented. The strategy has never come out of a model and it isn't going to."

The last sentence does the heavy lifting. It draws a line and tells the buyer exactly which side of it their money is buying.

Service business with a content operation. "We use AI to turn one long piece into the twelve shorter ones it should have been. The original thinking happens in the room with our team. The distribution is automated because distribution should be, which in our case means Buffer and a schedule."

This one reframes AI as a logistics decision rather than a creative one, which is both true for most operations and much easier for a customer to accept.

Solo operator, high volume. "I use AI heavily. It's how one person covers what usually takes three. What I don't do is publish anything I haven't checked against my own client work, which is why you'll occasionally see me contradict something I wrote last year."

The admission of contradiction is the credibility move. Nobody who's faking it volunteers that they've been wrong.

Technical or regulated work. "Drafting and documentation use AI assistance. Anything that carries a professional judgment is reviewed and signed by a named person, and we keep the working record on every engagement. Happy to walk you through what that record contains."

Formal register, same structure. The offer at the end is what separates it from boilerplate, and about one prospect in fifteen takes you up on it, which is one prospect in fifteen who now trusts you considerably more than they did.

Notice what all four have in common. They name the tool without apology, they're specific about the boundary, and the interesting half of the sentence is always what the human does. None of them use the word leverage.

I've framed all of this around trust rather than compliance on purpose, because the compliance framing gets you the worst version.

Under the EU AI Act, the duty to put a visible label on published content sits mostly with the organization publishing it, not the model provider. Anthropic embeds a machine readable mark. Turning that into something a human reader sees is a separate step and it's largely your side of the fence.

You could treat that as a box to tick. Add a badge, add a footer, done, minimum viable compliance.

Or you could notice that regulation just handed you a free excuse to say something true about how you work, at a moment when your competitors are either silent or slapping on badges, to an audience that has spent two years quietly assuming everything is machine made and finding no one willing to talk about it plainly.

That's not a compliance problem. That's an opening.

The businesses that come out of the next two years with the strongest trust position won't be the ones who avoided AI, because that number is heading toward zero and everyone knows it. They'll be the ones who were specific and unembarrassed about how they work while everybody else was hiding or bragging.

Write the line. Twenty minutes. It'll do more for your close rate than the next three blog posts.

"My clients will use it to negotiate me down."

Some will try. Here's the thing though: the client who says "you used AI so why am I paying this" was going to find a different reason if that one weren't available. Price pressure is not caused by disclosure, it's caused by a buyer who doesn't understand what they're buying, and vagueness makes that worse rather than better.

The counter is in the sentence itself. If your disclosure names a specific thing you do that they can't get elsewhere, you've answered the objection before it's raised. If your disclosure doesn't name anything specific, then the objection is fair and the problem isn't the disclosure.

That's uncomfortable and it's also the most useful thing this exercise surfaces. Writing the line honestly forces you to articulate your value, and if you can't, you've found something more important than a website update.

"Nobody's asking, so why bring it up."

Because the people who wonder mostly don't ask. That's the whole mechanism. Asking is awkward, it implies suspicion, and a prospect who's mildly unsure just goes quiet and picks somebody else. You never get the data point.

You are already paying for this. You just can't see the invoice.

The other reason: the moment when this becomes mandatory rather than optional is coming, and the businesses who wrote something considered in 2026 will look thoughtful. The ones scrambling to add a disclosure banner in 2027 because a platform started requiring it will look like they got caught. Same information, completely different read, entirely down to timing.

Getting there first costs you twenty minutes and buys you the good version of the story.

Jordan

The AI Newsroom is written by Jordan Hale. This issue contains affiliate links to tools I actually use. If you sign up through them I may earn a commission at no extra cost to you.